Last updated: 5 August 2026

Drafted by Charlie from what the project and the 5 August session established, rather than from a fresh interview. Lines marked [CONFIRM] are inferences. Correct them and delete the marker; everything unmarked is established fact from CLAUDE.md, the trip folders, or a decision Annette made out loud.

---

What this project is

Annette's trip planning workspace. One folder per trip holding the research, the itinerary, the budget, the bookings log and the packing list. Charlie does the destination research, drafts the plans, and publishes each finished one as a gated hub page.

The one thing it has to accomplish: produce a plan the people actually going can act on without redoing the research, and without being surprised when they arrive.

Scope is planning, never booking. Charlie researches, compares and recommends. Annette makes every purchase herself. No flights, hotels, cars, tours, deposits or holds, and no entering card or passport details anywhere. This is the rule most likely to be tested in a travel project, which is why it is stated twice in CLAUDE.md.

Live trips as of this date:

---

Primary audience

The reader is usually not Annette, and it changes per trip. This is the single most important thing about this project and it was learned the hard way on 5 August, when the Sweden report had to be rewritten from a report about three people into a note to them.

Three patterns so far:

1. A plan for people Annette isn't traveling with. Sweden is this: Bob, Brent and Krista are going, Annette is not. The report is written from her to the three of them, and two of them had never seen a word of it before opening the page.
2. A plan for Bob alone. Mont Blanc is this. He is an experienced climber going as a solo client, so the report assumes real competence and skips the basics.
3. A plan for a trip Annette is on. [CONFIRM] Alaska appears to be this, but the party was never confirmed and that is still an open question on that page.

So every deliverable has to name its reader before it is drafted. "Who opens this page, and what do they already know?" is the first question, not an afterthought. A reader who has never seen the trip discussed needs orienting in the first paragraph: what the trip is, that nothing is booked, and what is being asked of them.

What they know: these are capable adults, often more expert than Charlie in their own domain. Bob has real glaciated-peak experience. Bob, Brent and Krista have chartered big boats together for a week at a time, several times. Do not explain their own sport to them.

What they care about: whether the plan survives contact with reality. Prices that are still true, rules that actually apply to them, and the specific thing that will bite on arrival if nobody checked it.

---

Voice and tone

Written from Annette to the reader, first person. Not a consultant's report handed over, and not Charlie's voice. "I looked into this," "here's what I found," "tell me which you'd rather have."

Peer to peer, direct, plainspoken. Contractions throughout. Warm but not chatty. Honest about what is not known. Willing to say "don't" and then say why.

Hard rules, inherited from the global file:

• No em dashes. Anywhere, ever. Commas, colons, parentheses, new sentences.
• American English, except inside a direct quotation, a proper noun, or a vendor's own label. Quoted source text keeps its original spelling. This has come up: a Nautilus quote about "mostly harbour moorings" and a boat tester's "almost two metres" both stay British because altering a quotation is a misquote, not a spelling fix.
• Every URL is a clickable markdown link. Never a bare URL, never a local path.
• No sticky or fixed elements in anything published.

---

Authentic material

The specifics of the actual people are what make these plans useful, and they beat any generic advice. Pull them from annette.md and from what the travelers say, never assume.

Real examples that changed a deliverable:

• Brent is 6 foot 4. That single fact overturned the Sweden boat recommendation, because the boat previously recommended has 1.84 m of saloon headroom and he would have hit his head all week.
• Brent and Krista are married. That turned three cabins for three people into two, which changed the sleeping plan and made a cheaper two-cabin boat genuinely possible, which then had to be argued against on headroom grounds.
• Bob holds ASA 101, 103, 104 and 114. That answered the gating question on whether they could charter bareboat at all.
• The three of them have chartered big boats together, repeatedly. That is the half of a charter operator's requirement most applicants cannot answer, so it leads the email.
• Bob has real glaciated-peak experience, which is why the Mont Blanc report treats the climbing as the easy part and the refuge booking as the hard part.

The pattern: ask what is specifically true about these people, then let it change the recommendation rather than appending it as a note.

---

What NOT to do

Never book, hold, or reserve anything. Never enter card, passport, or Global Entry details. This is the project's defining constraint.

Never reference previous versions of a report. Annette's explicit instruction, twice on 5 August. Each deliverable reads as though written once, today. No "revised after," no old plan versus new plan, no keeping the superseded analysis "because the reasoning still matters," no internal pointers like "see section 3e." If a recommendation changes, the new document simply states the new recommendation.

Never state a price, a visa rule, an entry requirement or an opening time from memory. Search it, link it, and stamp it with the date checked. Anything older than about 30 days gets re-verified before anyone acts on it.

Never publish an itinerary ungated. Trip plans reveal when a house is empty. Every trip page uses the password gate and noindex.

Never commit confirmation numbers, passport numbers, frequent-flyer logins or card details. bookings.md is gitignored by default and stays that way.

Do not trust a manufacturer's own current brochure without a cross-check. Learned on the Sweden report: Bavaria's live CRUISER LINE brochure lists the Cruiser 37's headroom figure in the Cruiser 34's row. Checking only that document produces a 9 cm error on the one dimension that mattered.

Do not trust model names. "Bavaria 37 Cruiser" and "Bavaria Cruiser 37" are different boats by different designers. Brokers use the names interchangeably. Verify by dimensions and designer.

Do not assume a published operator itinerary is compatible with the operator's own contract. Every published Stockholm week-long route ends in central Stockholm, and the charter terms require the boat back at base the night before.

---

Output formats

The deliverable is a gated hub page, generated from markdown by the canonical ~/.claude/skills/publish-to-hub/scripts/build_hub_page.py. Edit the markdown and rerun the builder. Never hand-edit the generated .astro; it is overwritten every build.

Front matter carries title, title_line2, hero, hero_alt and description.

Trip folder structure, from _templates/trip/. Copy the folder, do not write from scratch:

trips/<year>-<destination-slug>/
  README.md        overview, dates, who's going, status
  research.md      destination notes, sources, date-stamped
  itinerary.md     day by day
  budget.md        estimates vs actuals
  bookings.md      gitignored, never published
  packing.md       trip-specific packing

Deep working files can stay local, but only if Annette says so. The Sweden trip produced a 968-line island-by-island companion (where-to-explore.md) that is genuinely useful and is sitting unread in a local file, which is exactly the shape of thing she does not read. When in doubt, publish it as its own gated page.

---

Stakeholders

The travelers themselves are the reviewers, and they read the page directly. That is unusual for Annette's projects and it changes the writing: the page has to survive being read by the person it is about.

[CONFIRM] Nobody else reviews these before they go out.

Redaction rules still apply in full. Other people's health information never gets written down here. Annette's own health story is public by her standing choice, but a travel plan rarely needs it. Trip participants' names are fine on a gated page.

---

Success looks like

The people going can act on the plan without redoing the research:

• They know what to book, from whom, at what price, and by when.
• They know what will bite them, and it is stated plainly rather than buried.
• Every price and rule carries the date it was checked, so they can tell what needs re-verifying.
• What is not known is listed explicitly rather than glossed over.
• Nobody arrives and discovers something that a competent planner would have caught.

The counter-example is the whole point: a boat that fits everyone except the tallest person aboard is a failure, and it would only have shown up at check-in.

---

Key terminology and constraints

• Charlie never books. Annette purchases.
• Date-stamp everything perishable. Airfares, hotel rates, visa rules, entry requirements, opening hours, harbor fees.
• Re-verify anything older than about 30 days before acting on it.
• Gate every itinerary, and never submit any of these pages to a search engine.
• bookings.md, PDFs and anything under documents/ are gitignored.
• Names to get right: Bob, Brent, Krista. [CONFIRM] Bob's surname is Manthy; Brent and Krista's surname is unknown to Charlie.
• Trip-relevant personal facts live in annette.md, not here. Pull health, energy, food preferences and travel companions from there rather than assuming.

---

Open questions for Annette

1. Alaska: who is going, and is that trip still live at all? Its fares are 2026 and the August dates have passed.
2. Is Annette on any of these trips, or is this workspace mostly for planning other people's? It changes who each report is addressed to.
3. Should the Sweden companion file (where-to-explore.md) be published as its own gated hub page?
4. costs.md on the Sweden trip is now inconsistent with the main report: it assumes five crew and predates the boat change. Update it or drop the link?

---

Drafted by the project-context skill, 5 August 2026. Update by editing this file directly or by running /project-context again.